10/25Gbit/s Ethernet MAC/PCS
The 10/25GMAC IP core is a Low-Latency Ethernet MAC with a latency of 44.8ns in 2749 LUTs for 10Gbit/s and 20.5ns in 2680 LUTs for 25Gbit/s. When combined with the Low-Latency 10/25GPCS, the full packet round trip time for 10Gbit/s is ( MAC Input -> Wire -> MAC Output ) 153.8ns in 5153 LUTs; 25Gbit/s ( MAC Input -> Wire -> MAC Output ) is 128ns in 7930 LUTs.
The IEEE 802.3by compliant 10/25G MAC/PCS was designed in house at Chevin Technology, to provide an easy path to the integration of protocols such as TCP/IP and UDP protocols in your FPGA, whilst using minimal FPGA resources. The 10/25GMAC simplifies the synthesis of ultra-fast Duplex 25Gbit/s Ethernet for FPGAs.
Chevin Technology offer a detailed user guide, expert support and design services to assist in the implementation of 10/25Gbit/s Ethernet connectivity in Xilinx Virtex® UltraScale™ FPGAs. A reference design is available for technology partner Alpha Data’s ADM-PCIE-8V3, ADM-PCE-9V3 boards. The 10/25G MAC/PCS is also compatible with Xilinx’s VCU108 board and has been tested with the Mellanox MCX4121A-ACAT NIC. Flexible licensing terms are available with Chevin Technology IP cores, to allow for the unique requirements of each customers’ project.
We understand that efficiency and reliability are crucial to our customers, and have created a powerful CRC32 checker & generator engine that checks the TX and RX data for errors, on a 64bit wide bus @ 390.625MHz.
Latency of the 10/25GMAC can be reduced even further by the use of Cut-through mode; the first byte appears only 8 nanoseconds after arriving at 25GMII. Alternatively, the Store-and-Forward mode reduces application workload, as the 25GMAC drops all corrupt frames. The Frame Checksum verifies frame integrity; the CRC32 check result is available 8 nanoseconds after the final byte is received.
The Deficit Idle Count optimizes the Inter Frame Gap (IFG) for absolute maximum Throughput and minimum Latency by maintaining an average IFG count.